CN111445473B - Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction - Google Patents
Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction Download PDFInfo
- Publication number
- CN111445473B CN111445473B CN202010243718.9A CN202010243718A CN111445473B CN 111445473 B CN111445473 B CN 111445473B CN 202010243718 A CN202010243718 A CN 202010243718A CN 111445473 B CN111445473 B CN 111445473B
- Authority
- CN
- China
- Prior art keywords
- intravascular ultrasound
- image
- cross
- segmentation result
- longitudinal axis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20068—Projection on vertical or horizontal image axis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
A vascular membrane accurate segmentation method and system based on multi-angle reconstruction of an intravascular ultrasound image sequence are provided, a corresponding intravascular ultrasound longitudinal axis image is obtained by reconstructing an intravascular ultrasound cross-section image sequence, then an improved clustering method is adopted to perform preliminary segmentation on vascular membranes in the intravascular ultrasound longitudinal axis image, and then the vascular membrane accurate segmentation result under the cross-section image is recovered according to the preliminary segmentation result. The invention can obtain the vascular membrane segmentation results of two image modes (namely a cross-sectional image mode and a longitudinal axis image mode) of intravascular ultrasound at the same time, and overcomes the influence of bifurcation and bypass blood vessels to a certain extent. Comprehensively analyzing the blood vessel condition from the cross-section view and the longitudinal axis view, calculating quantitative parameters, more comprehensively evaluating the blood vessel condition, providing reference for further diagnosis of doctors and having clinical practical value.
Description
Technical Field
The invention relates to a technology in the field of image processing, in particular to a vascular membrane precise segmentation method based on multi-angle reconstruction of an intravascular ultrasound image sequence.
Background
The intravascular ultrasound, as shown in figure 1, is carried out by sending an ultrasound probe into a blood vessel cavity for imaging through a catheter technology, and can observe the shape of the lumen and the structure of the wall of the blood vessel through the image of the cross section of the blood vessel, has the advantages of intuitionism, accuracy and the like, is considered as a gold standard for diagnosing coronary heart disease, and plays a very important role in improving the recognition of coronary artery lesions and guiding interventional therapy. By accurately dividing the boundaries of vascular membranes (medium-adventitia and intima), the reference lumen and the actual lumen are determined, quantitative analysis of vascular conditions by quantitative parameters (lumen cross-sectional area, lumen area stenosis, etc.) can be calculated, and the severity of lesions can be judged.
The existing manual segmentation method manually depicts vascular membranes (medium-adventitia and intima), the effect of the method is different from person to person and the process is time-consuming, so that the vascular membrane automatic segmentation algorithm is worthy of research and realization. Because intravascular ultrasound is directly acquired as a cross-sectional image, most of the existing automatic segmentation algorithms for vascular membranes directly process the cross-sectional image. However, one drawback of intravascular ultrasound is that the acquired cross-sectional images reflect only the vascular condition of a certain cross-section. The cross-sectional images are directly processed, and vascular structure continuity information contained in the whole intravascular ultrasound image sequence is ignored. In addition, when the cross-sectional image is processed, only the blood vessel structure information of the cross section and the adjacent cross sections can be utilized, so that the blood vessel is extremely easy to be influenced by blood vessel bifurcation and bypass blood vessels, and the accuracy of dividing blood vessel membranes is reduced.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a vascular membrane accurate segmentation method and a vascular membrane accurate segmentation system based on multi-angle reconstruction of an intravascular ultrasound image sequence, which are characterized in that vascular membrane segmentation results of a cross-section image sequence are restored after longitudinal axis images are segmented by introducing vascular structure continuity and starting from longitudinal axis mode image reconstruction, so that vascular membrane segmentation results of intravascular ultrasound image modes (namely a cross-section image mode and a longitudinal axis image mode) can be obtained simultaneously, vascular conditions are comprehensively analyzed from a cross-section view angle and a longitudinal axis view angle, quantitative parameters are calculated, the vascular conditions are more comprehensively evaluated, and a reference is provided for further diagnosis of doctors, and the method has clinical practical values.
The invention is realized by the following technical scheme:
the invention relates to a vascular membrane accurate segmentation method based on multi-angle reconstruction of an intravascular ultrasound image sequence, which comprises the steps of reconstructing an intravascular ultrasound cross-section image sequence to obtain a corresponding intravascular ultrasound longitudinal axis image, performing preliminary segmentation on vascular membranes in the intravascular ultrasound longitudinal axis image by adopting an improved clustering method, and recovering the vascular membrane accurate segmentation result under the cross-section image according to the preliminary segmentation result.
The intravascular ultrasound cross-section image sequence is a complete intravascular ultrasound cross-section image sequence.
The reconstruction is carried out by introducing continuity of a blood vessel structure contained in the whole cross-section image sequence and reflecting the continuity on a corresponding longitudinal axis image through sampling and interpolation technology; the specific operation comprises the following steps: resampling the intravascular ultrasound cross-section image sequence frame by frame, namely sampling the intravascular ultrasound cross-section image sequence by the pixel point in the center of the image, sampling the intravascular ultrasound cross-section image sequence angle by angle and diameter, interpolating all the sampling values of the cross-section images in the sequence at the same angle together, and reconstructing the longitudinal axis mode image corresponding to the angle.
The primary segmentation is carried out, and a blood vessel membrane rough segmentation result is obtained by sequentially using fuzzy clustering on the reconstructed intravascular ultrasound longitudinal axis image; then optimizing by a morphological filter to obtain a vascular membrane preliminary segmentation result of the longitudinal intravascular ultrasound image as an initial contour; and then the longitudinal axis image is used for calculating an external force field, and the preliminary segmentation result of the intravascular ultrasound longitudinal axis image vascular membrane is obtained through the optimization of the active contour model of a plurality of iterations.
The recovery means: obtaining a vascular membrane segmentation result of the cross-section image through the inverse process of reconstructing the longitudinal axis image; then evenly downsampling on the cross-sectional image to reduce the interference of bifurcation and accompanying bypass blood vessels; and finally, carrying out circle fitting smoothing on the downsampling result and optimizing the movable contour model to obtain a blood vessel membrane accurate segmentation result of the intravascular ultrasound cross-section image.
The invention relates to a system for realizing the method, which comprises the following steps: the device comprises a reconstruction unit, a preliminary segmentation unit, a recovery unit and an optimized segmentation unit, wherein: the reconstruction unit obtains a corresponding intravascular ultrasound longitudinal axis image sequence through frame-by-frame resampling and interpolation processing according to the complete intravascular ultrasound cross section image sequence; the preliminary segmentation unit adopts an improved clustering method to carry out preliminary segmentation on the intravascular ultrasound longitudinal axis image and obtain a segmentation result; the recovery unit extracts a blood vessel membrane preliminary segmentation result under the cross-section image from the preliminary segmentation result of the longitudinal axis image through an inverse reconstruction process; the optimized segmentation unit optimizes the preliminary segmentation result of the blood vessel membrane of the cross-section image through downsampling, fitting and smoothing treatment and the movable contour model so as to obtain an accurate segmentation result.
Technical effects
The invention integrally solves the problems that the cross-section image in the prior art cannot introduce continuity of a blood vessel structure and the interference exists in the longitudinal axis image. The longitudinal axis image sequence is obtained by reconstructing the cross-section image sequence, so that vascular structure continuity information is introduced, further, the influence of vascular bifurcation and accompanying bypass blood vessels is effectively weakened when the interference of the vascular bifurcation and the accompanying bypass blood vessels occurs, and a cross-section mode image, a longitudinal axis mode image and a more effective vascular membrane segmentation result are realized.
Drawings
FIG. 1 is a schematic diagram of the principle of intravascular ultrasound;
FIG. 2a is a cross-sectional intravascular ultrasound image of an embodiment, and FIG. 2b is an exemplary graph of the results of reconstructing a longitudinal intravascular ultrasound image;
FIG. 3a is an exemplary graph of the interference of the bypass blood vessel in the vertical axis image, and FIG. 3b is an exemplary graph of the segmentation result of the blood vessel membrane of the reconstructed cross-sectional image after the interference of the bypass blood vessel in the cross-sectional image and the downsampling recovery;
FIG. 4 is a flow chart of the method of the present invention;
in the figure: (a) is a cross-section intravascular ultrasound image sequence, (b) is a longitudinal axis intravascular ultrasound image reconstructed, (c) is subjected to fuzzy clustering to obtain a vascular membrane preliminary segmentation result, (d) is subjected to morphological filtering treatment to obtain a vascular membrane preliminary segmentation result, (e) is a longitudinal axis intravascular ultrasound image vascular membrane segmentation result, and (f) is subjected to downsampling to recover the cross-section intravascular ultrasound image vascular membrane segmentation result;
FIG. 5.1 is an exemplary diagram of a reconstruction unit, and FIG. 5.2 is an exemplary diagram of a recovery unit;
in the figure: the solid line is the sampling (interpolation) position, and the arrow indicates the sampling (interpolation) direction;
FIG. 6 is an exemplary diagram of an optimized segmentation unit;
FIG. 7 is an exemplary view of a longitudinal axis intravascular ultrasound image vessel membrane segmentation of an embodiment of the disclosed dataset;
in the figure: the dashed line represents the automatic segmentation result, and the solid line represents the gold standard;
FIG. 8 is an exemplary graph of cross-sectional intravascular ultrasound image vessel membrane segmentation in an embodiment of the disclosed dataset;
in the figure: the dashed line represents the automatic segmentation result, and the solid line represents the gold standard;
FIG. 9 is a longitudinal intravascular ultrasound image of the vascular membrane in a self-established dataset embodiment;
in the figure: the dashed line represents the automatic segmentation result, and the solid line represents the gold standard;
FIG. 10 is a longitudinal intravascular ultrasound image of a vascular membrane in a self-constructed dataset embodiment;
in the figure: the dashed line represents the automatic segmentation result and the solid line the gold standard.
Detailed Description
As shown in fig. 4, this embodiment relates to a method for precisely segmenting a vascular membrane based on multi-angle reconstruction of an intravascular ultrasound image sequence, which comprises reconstructing an intravascular ultrasound cross-sectional image sequence to obtain a corresponding intravascular ultrasound longitudinal axis image, performing preliminary segmentation on the vascular membrane in the intravascular ultrasound longitudinal axis image by using an improved clustering method, and recovering a precise segmentation result of the vascular membrane in the cross-sectional image according to the preliminary segmentation result.
The intravascular ultrasound cross-section image sequence is a complete intravascular ultrasound cross-section image sequence.
The reconstruction is carried out by introducing continuity of a blood vessel structure contained in the whole cross-section image sequence and reflecting the continuity on a corresponding longitudinal axis image through sampling and interpolation technology; the specific operation comprises the following steps: resampling the intravascular ultrasound cross-section image sequence frame by frame, namely sampling the intravascular ultrasound cross-section image sequence by the pixel point in the center of the image, sampling the intravascular ultrasound cross-section image sequence angle by angle and diameter, interpolating all the sampling values of the cross-section images in the sequence at the same angle together, and reconstructing the longitudinal axis mode image corresponding to the angle.
A 250 frame (cross-sectional) intravascular ultrasound image sequence of 384 pixels x 384 pixels is illustrated. The size of the image sequence is 384×384×250, firstly, the image sequence is sampled along the radial direction (the center of the over-image is the pixel point on 192 th row and 192 th column, the length is 384 pixels) and the angle is sampled every 1 degree (180 times in total), and the image sequence is converted into an intravascular ultrasound image sequence in a polar coordinate system, and the size is 384×180×250. Then, the intravascular ultrasound image sequences in the polar coordinate system (for example, the first column of all the images forms a longitudinal intravascular ultrasound image of 0 °) are interpolated column by column to obtain corresponding longitudinal intravascular ultrasound image sequences (the size is 384×250×180), as shown in fig. 4.
The primary segmentation is carried out, and a blood vessel membrane rough segmentation result is obtained by using fuzzy clustering on the reconstructed intravascular ultrasound longitudinal axis image; then optimizing by a morphological filter to obtain a vascular membrane preliminary segmentation result of the longitudinal intravascular ultrasound image as an initial contour; and then the longitudinal axis image is used for calculating an external force field, and the preliminary segmentation result of the intravascular ultrasound longitudinal axis image vascular membrane is obtained through the optimization of the active contour model of a plurality of iterations, and the specific steps comprise:
(1) for longitudinal intravascular ultrasound images, first a rough vessel membrane segmentation result is obtained using fuzzy clustering, as shown in fig. 4 (c).
The fuzzy clustering, namely the clustering algorithm based on division, enables the similarity between objects divided into the same cluster to be the largest, and the similarity between different clusters to be the smallest. The objective function is:
wherein: u= [ U ] ik ]As a membership matrix, u ik Is the membership of the kth sample to the ith class, d ik 2 ＝||x k -v i || 2 Is sample x k And cluster center (mean value))v i Is a euclidean distance of (c). The constraint is that the sum of membership degrees of a certain sample to each cluster is 1, namely:
the clustering problem translates into:
The fuzzy clustering method comprises the following specific algorithm steps:
i) Setting the clustering number c and the parameter m;
ii) initializing a membership matrix U (0), wherein the sum of elements in each column of U (0) is 1;
iii) Calculating a new cluster center Vj;
iv) calculating a new membership matrix U;
v) stopping the iteration when the termination condition U (k+1) -U (k) is smaller than or equal to e, otherwise continuing.
(2) In order to reduce the fine interference due to the speckle noise in fig. 4 (c), an open operation noise reduction is performed using a circular morphological filter having a diameter of 9 pixels.
Since the interference suffered by the longitudinal axis intravascular ultrasound image, especially the interference from the bifurcation and the bypass blood vessel, often occurs in a limited and continuous angle, when the segmentation result of the blood vessel membrane of the (cross section) intravascular ultrasound image is restored, if the segmentation result of the blood vessel membrane of the longitudinal axis intravascular ultrasound image is directly used for restoration to obtain the segmentation result of the blood vessel membrane of the final cross section image, the situation as shown in fig. 3 (b) and fig. 6 (a) is generated, and the segmentation failure is caused. In order to effectively overcome such a situation, the recovery means of this embodiment is as follows: as shown in fig. 5.2, the vessel membrane segmentation result of the cross-sectional image is obtained by the inverse process of reconstructing the vertical axis image; as shown in fig. 6, the cross-sectional image is then uniformly downsampled to reduce the interference of bifurcated, bypass vessels; finally, the downsampling result circle fitting smoothness and the movable contour model are optimized to obtain the accurate segmentation result of the blood vessel membrane of the intravascular ultrasound cross section image, which is specifically as follows: after the vascular membrane segmentation result of the (cross section) intravascular ultrasound image is restored by using the longitudinal intravascular ultrasound image with all angles, uniform downsampling with the interval of 40 degrees is performed in the radial direction, and then the downsampling result is subjected to circle fitting to obtain the segmentation result. In order to better segment the lumen contour with irregular shape, a movable contour model is introduced at last to optimize, and a final cross-section image vascular membrane segmentation result is obtained. In the invention, a snake model is used, the iteration times are only 15 times, and the calculation load is very small.
The segmentation method proposed in this embodiment is subjected to an actual intravascular ultrasound image test. The data set used comprised two parts, the first part being a public data set, 10 patients were acquired for a total of 2175 frames of images, wherein 7 patients were acquired for 250 frames, 2 patients were acquired for 150 frames, and 1 patient was acquired for 125 frames, the gold standard corresponding to the segmentation of the vascular membrane was noted by four experienced clinicians (the original gold standard of the data set was noted by the doctor every 5 frames, and a total of 435 images contained vascular membrane markers; the remaining images were generated by volunteers based on the existing gold standard prior to implementation, under clinician guidance, the gold standard of the vertical axis images was combined with the original and supplemental gold standard), these images contained typical disturbances of bifurcation, bypass vessels, etc. The collection device used was Si5 (Volco Co.) equipped with a 20MHz Eagle Eye catheter. The second part is a self-built data set acquired by the department of cardiology of a Zhongshan hospital affiliated to Shanghai's complex denier university, and acquires 5370 frames of images of 2 patients, wherein 1 patient is acquired twice in different periods, 3562 frames (patient 1-1) are acquired for the first time, 806 frames (patient 1-2) are acquired for the second time, and 1002 frames (patient 2) are acquired for the other patient. This data was labeled with only the corresponding mid-adventitial vessels, labeled by two volunteers and modified multiple times under the direction of the cardiologist. The collection device used was iLab (Boston Scientific company) equipped with a 40MHz optical catheter.
Three evaluation indexes are calculated based on the segmentation result and the corresponding gold standard: huo Siduo f distance (Hausdorff distance, HD), jacked measure (Jaccard measure index, JM) and area difference percentage (Percentage of area difference, PAD) to quantitatively evaluate the degree of similarity between the segmentation result and the gold standard.
HD is defined as the maximum distance of each point in the segmentation result contour to its nearest neighbor of the corresponding gold standard:
wherein: b (B) auto For the contour of the segmentation result, B man For the contour corresponding to the gold standard, dist is the Euclidean distance between points p and q.
JM is defined as the intersection of the segmented result and the gold standard divided by the union of the two to measure the similarity between the segmented result and the gold standard:
wherein: z is Z auto To divide the resulting region, Z man Area is the Area corresponding to the gold standard.
PAD is defined as the percentage difference between the difference set of the segmentation result and the gold standard, and is used to measure the difference between the segmentation result and the gold standard:
smaller HD and PAD, and larger JM means that the segmentation results are closer to the gold standard.
In the present embodiment, as shown in fig. 2 (b), the lumen region in the longitudinal-axis intravascular ultrasound image reconstructed based on the cross-sectional intravascular ultrasound image sequence is clearly visible, and the vascular condition in the longitudinal axis direction can be effectively displayed. As shown in fig. 3 and 6, the downsampling recovery strategy is effective in reducing the interference caused by bifurcation and bypass vessels.
When the acquired vascular membrane ultrasonic cross-section image sequence is overlong, the size of the longitudinal axis image obtained by reconstruction is increased, and an efficient segmentation algorithm is also required to be introduced from the viewpoint of time consumption of calculation. After the fuzzy clustering result of the vertical axis image, namely the initial segmentation result of the vascular membrane, some tiny interference as shown in fig. 4 (c) exists, and optimization is needed. These small interference areas are smaller and can be effectively distinguished from the areas formed by the outer membrane structures by the size of the areas. Therefore, the small interference can be effectively removed by performing an open operation with a morphological filter of a suitable size, as shown in fig. 4 (d).
As shown in fig. 7, 8 and 9, 10, for the vessel membrane segmentation results and corresponding gold standard of the longitudinal and cross-sectional intravascular ultrasound images of the public and self-built datasets, it can be observed that the automatic segmentation results are very close to the gold standard.
From the evaluation index point of view, as shown in table 1, for the public dataset, the evaluation index for mid-adventitia and intima segmentation was comparable or superior to the other three published methods. In particular, in table 2, the evaluation index of the present algorithm is better for intravascular ultrasound images with bifurcated, bypass vessels than for the other three methods. In addition, compared with other three methods, the method realizes the segmentation of the blood vessel membrane of the cross section and the longitudinal axis images at the same time, and has obvious advantages for evaluating the condition of the whole blood vessel compared with other three methods which only realize the segmentation of the blood vessel membrane of the cross section images. For the self-built data set, as shown in table 3, more effective mid-adventitia segmentation is realized for the 3-segment intravascular ultrasound image sequences of 2 patients, which shows that the method is not easily affected by a certain change of the sequence length.
From the point of view of computational time, as shown in table 4, the computational time of the present algorithm is advantageous over most other automatic or semi-automatic methods.
Table 1.
Table 2.
TABLE 3 Table 3
TABLE 4 Table 4
In summary, the embodiment acquires a complete (cross-section) intravascular ultrasound image sequence, and reconstructs the corresponding longitudinal intravascular ultrasound image sequence through image processing. On the basis, the embodiment can realize the full-automatic segmentation of the vascular membrane of the intravascular ultrasound image with two modes of a longitudinal axis and a cross section, has higher automaticity, robustness and accuracy, and especially realizes better segmentation effect for the intravascular ultrasound image with bifurcation and bypass blood vessels.
The foregoing embodiments may be partially modified in numerous ways by those skilled in the art without departing from the principles and spirit of the invention, the scope of which is defined in the claims and not by the foregoing embodiments, and all such implementations are within the scope of the invention.
Claims (6)
1. A vascular membrane accurate segmentation method based on multi-angle reconstruction of an intravascular ultrasound image sequence is characterized in that a corresponding intravascular ultrasound longitudinal axis image is obtained by reconstructing an intravascular ultrasound cross-section image sequence, then an improved clustering method is adopted to perform primary segmentation on vascular membranes in the intravascular ultrasound longitudinal axis image, and then the vascular membrane accurate segmentation result under the cross-section image is recovered according to the primary segmentation result;
the intravascular ultrasound cross-section image sequence is a complete intravascular ultrasound cross-section image sequence;
the primary segmentation is carried out, and a fuzzy clustering method is used for the reconstructed intravascular ultrasound longitudinal axis image to obtain a blood vessel membrane rough segmentation result; then optimizing by a morphological filter to obtain a vascular membrane preliminary segmentation result of the longitudinal intravascular ultrasound image as an initial contour; and then the longitudinal axis image is used for calculating an external force field, and the preliminary segmentation result of the intravascular ultrasound longitudinal axis image vascular membrane is obtained through the optimization of the active contour model of a plurality of iterations, and the specific steps comprise:
(1) for a longitudinal intravascular ultrasound image, firstly, fuzzy clustering is used to obtain a rough vascular membrane segmentation result;
(2) performing open operation noise reduction optimization by adopting a circular morphological filter with the diameter of 9 pixels;
the fuzzy clustering, namely based on a clustering algorithm of division, makes the similarity among objects divided into the same cluster maximum, and the similarity among different clusters minimum, and the objective function is as follows:
wherein: u= [ U ] ik ]As a membership matrix, u ik Is the membership of the kth sample to the ith class, d ik 2 ＝||x k -v i || 2 Is sample x k And cluster center v i C is the number of clusters, m is a parameter, and the constraint condition is that the sum of membership degrees of a certain sample to each cluster is 1, namely: />
The clustering problem translates into:
2. The method of claim 1, wherein the reconstructing is reflected on the corresponding vertical axis image by introducing continuity of the vascular structure contained in the whole cross-sectional image sequence and by sampling and interpolation techniques, and specifically comprises: resampling the intravascular ultrasound cross-section image sequence frame by frame, namely sampling the intravascular ultrasound cross-section image sequence by the pixel point in the center of the image, sampling the intravascular ultrasound cross-section image sequence angle by angle and diameter, interpolating all the sampling values of the cross-section images in the sequence at the same angle together, and reconstructing the longitudinal axis images corresponding to the angles.
3. The method for precisely segmenting the blood vessel membrane according to claim 1, wherein the recovery means: obtaining a vascular membrane segmentation result of the cross-section image through the inverse process of reconstructing the longitudinal axis image; uniformly downsampling on the cross-sectional image to reduce bifurcation and bypass vascular interference; and finally, carrying out circle fitting smoothing on the downsampling result and optimizing the movable contour model to obtain a blood vessel membrane accurate segmentation result of the intravascular ultrasound cross-section image.
4. The method for precisely segmenting the blood vessel membrane according to claim 1, wherein the fuzzy clustering comprises the following specific algorithm steps:
i) Setting the clustering number c and the parameter m;
ii) initializing a membership matrix U (0), wherein the sum of elements in each column of U (0) is 1;
iii) Calculating a new cluster center Vj;
iv) calculating a new membership matrix U;
v) stopping the iteration when the termination condition U (k+1) -U (k) is smaller than or equal to e, otherwise continuing.
5. A method for precisely segmenting a vascular membrane according to claim 1 or 3, wherein the recovering comprises: after restoring the vascular membrane segmentation result of the cross-section intravascular ultrasound image by using the longitudinal intravascular ultrasound image of all angles, uniformly downsampling at intervals of 40 degrees in the radial direction, and performing circle fitting on the downsampling result to obtain the segmentation result; and finally, introducing a movable contour model for optimization to obtain a final segmentation result of the blood vessel membrane of the cross-section image.
6. A system implementing the method of any one of claims 1-5, comprising: the device comprises a reconstruction unit, a preliminary segmentation unit, a recovery unit and an optimized segmentation unit, wherein: the reconstruction unit obtains a corresponding intravascular ultrasound longitudinal axis image through frame-by-frame resampling according to the complete intravascular ultrasound cross section image sequence, and the preliminary segmentation unit carries out preliminary segmentation on the intravascular ultrasound longitudinal axis image by adopting an improved clustering method and obtains a segmentation result; the recovery unit extracts a blood vessel membrane preliminary segmentation result under the cross-section image from the preliminary segmentation result of the longitudinal axis image through an inverse reconstruction process; the optimized segmentation unit optimizes the preliminary segmentation result of the blood vessel membrane of the cross-section image through downsampling, fitting and smoothing treatment and the movable contour model so as to obtain an accurate segmentation result;
the primary segmentation is carried out, and a fuzzy clustering method is used for the reconstructed intravascular ultrasound longitudinal axis image to obtain a blood vessel membrane rough segmentation result; then optimizing by a morphological filter to obtain a vascular membrane preliminary segmentation result of the longitudinal intravascular ultrasound image as an initial contour; and then the longitudinal axis image is used for calculating an external force field, and the preliminary segmentation result of the intravascular ultrasound longitudinal axis image vascular membrane is obtained through the optimization of the active contour model of a plurality of iterations, and the specific steps comprise:
(1) for a longitudinal intravascular ultrasound image, firstly, fuzzy clustering is used to obtain a rough vascular membrane segmentation result;
(2) performing open operation noise reduction optimization by adopting a circular morphological filter with the diameter of 9 pixels;
the fuzzy clustering, namely based on a clustering algorithm of division, makes the similarity among objects divided into the same cluster maximum, and the similarity among different clusters minimum, and the objective function is as follows:
wherein: u= [ U ] ik ]As a membership matrix, u ik Is the membership of the kth sample to the ith class, d ik 2 ＝||x k -v i || 2 Is sample x k And cluster center v i C is the number of clusters, m is a parameter, and the constraint condition is that the sum of membership degrees of a certain sample to each cluster is 1, namely: />
The clustering problem translates into:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010243718.9A CN111445473B (en) | 2020-03-31 | 2020-03-31 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010243718.9A CN111445473B (en) | 2020-03-31 | 2020-03-31 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111445473A CN111445473A (en) | 2020-07-24 |
| CN111445473B true CN111445473B (en) | 2023-04-28 |
Family
ID=71652595
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010243718.9A Active CN111445473B (en) | 2020-03-31 | 2020-03-31 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111445473B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112927212B (en) * | 2021-03-11 | 2023-10-27 | 上海移视网络科技有限公司 | OCT cardiovascular plaque automatic identification and analysis method based on deep learning |
| CN114881948B (en) * | 2022-04-26 | 2023-05-12 | 青岛埃米博创医疗科技有限公司 | Monte Carlo algorithm-based method for automatically generating vascular teaching aid by exploring box |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2449080A1 (en) * | 2003-11-13 | 2005-05-13 | Centre Hospitalier De L'universite De Montreal - Chum | Apparatus and method for intravascular ultrasound image segmentation: a fast-marching method |
| CN104376549A (en) * | 2014-11-20 | 2015-02-25 | 华北电力大学（保定） | Intravascular ultrasound image and intravascular-OCT image fusing method |
| CN108335304A (en) * | 2018-02-07 | 2018-07-27 | 华侨大学 | A kind of aortic aneurysm dividing method of abdominal CT scan sequence image |
| CN109166083A (en) * | 2018-09-03 | 2019-01-08 | 哈尔滨工业大学 | A method of for removing underwater picture bubble noise |
| CN109345538A (en) * | 2018-08-30 | 2019-02-15 | 华南理工大学 | A kind of Segmentation Method of Retinal Blood Vessels based on convolutional neural networks |
| CN110223271A (en) * | 2019-04-30 | 2019-09-10 | 深圳市阅影科技有限公司 | The automatic horizontal collection dividing method and device of blood-vessel image |
| CN110889846A (en) * | 2019-12-03 | 2020-03-17 | 哈尔滨理工大学 | Diabetes retina image optic disk segmentation method based on FCM |
-
2020
- 2020-03-31 CN CN202010243718.9A patent/CN111445473B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2449080A1 (en) * | 2003-11-13 | 2005-05-13 | Centre Hospitalier De L'universite De Montreal - Chum | Apparatus and method for intravascular ultrasound image segmentation: a fast-marching method |
| CN104376549A (en) * | 2014-11-20 | 2015-02-25 | 华北电力大学（保定） | Intravascular ultrasound image and intravascular-OCT image fusing method |
| CN108335304A (en) * | 2018-02-07 | 2018-07-27 | 华侨大学 | A kind of aortic aneurysm dividing method of abdominal CT scan sequence image |
| CN109345538A (en) * | 2018-08-30 | 2019-02-15 | 华南理工大学 | A kind of Segmentation Method of Retinal Blood Vessels based on convolutional neural networks |
| CN109166083A (en) * | 2018-09-03 | 2019-01-08 | 哈尔滨工业大学 | A method of for removing underwater picture bubble noise |
| CN110223271A (en) * | 2019-04-30 | 2019-09-10 | 深圳市阅影科技有限公司 | The automatic horizontal collection dividing method and device of blood-vessel image |
| CN110889846A (en) * | 2019-12-03 | 2020-03-17 | 哈尔滨理工大学 | Diabetes retina image optic disk segmentation method based on FCM |
Non-Patent Citations (2)
| Title |
|---|
| 任子晖等.基于Curvelet变换与遗传神经网络的血管分割算法.《计算机应用与软件》.2019,全文. * |
| 方玉宏等.肾脏结构与功能的光学监测研究进展.《福建师范大学学报》.2018,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111445473A (en) | 2020-07-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pellot et al. | A 3D reconstruction of vascular structures from two X-ray angiograms using an adapted simulated annealing algorithm | |
| US7978916B2 (en) | System and method for identifying a vascular border | |
| Zhou et al. | The detection and quantification of retinopathy using digital angiograms | |
| Schmitt et al. | New methods for the computer-assisted 3-D reconstruction of neurons from confocal image stacks | |
| CN110448319B (en) | Blood flow velocity calculation method based on contrast image and coronary artery | |
| Qian et al. | An integrated method for atherosclerotic carotid plaque segmentation in ultrasound image | |
| CN113420826B (en) | Liver focus image processing system and image processing method | |
| CN111445473B (en) | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction | |
| WO2012126070A1 (en) | Automatic volumetric analysis and 3d registration of cross sectional oct images of a stent in a body vessel | |
| CN113470137B (en) | IVOCT image guide wire artifact removing method based on gray-scale weighting | |
| CN112837306B (en) | Coronary artery disease lesion functional quantitative method based on deep learning and mesopic theory | |
| WO2004017823A2 (en) | System and method for identifying a vascular border | |
| Maaliw et al. | An enhanced segmentation and deep learning architecture for early diabetic retinopathy detection | |
| Huang et al. | Automatic retinal vessel segmentation based on an improved U-Net approach | |
| CN113470060B (en) | Coronary artery multi-angle curved surface reconstruction visualization method based on CT image | |
| Acosta-Mesa et al. | Cervical cancer detection using colposcopic images: a temporal approach | |
| CN113592802B (en) | Mitral valve annular displacement automatic detection system based on ultrasonic image | |
| WO2022096867A1 (en) | Image processing of intravascular ultrasound images | |
| Subramaniam et al. | Measurement of Intima-Media Thickness Depending on Intima Media Complex Segmentation by Deep Neural Networks | |
| CN114419015A (en) | Brain function fusion analysis method based on multi-modal registration | |
| CN114343693A (en) | Aortic dissection diagnosis method and device | |
| CN115249248A (en) | Retinal artery and vein blood vessel direct identification method and system based on fundus image | |
| Zhan et al. | Recognition of angiographic atherosclerotic plaque development based on deep learning | |
| CN117809839B (en) | Correlation analysis method for predicting hypertensive retinopathy and related factors | |
| Cai et al. | Detection of 3D Arterial Centerline Extraction in Spiral CT Coronary Angiography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |